1/7 The difficulty retarget algorithm in Bitcoin is terrible. It fails to achieve its stated goal - 2016 blocks every 2 weeks - in several ways:
2/7 1. It's not balanced over long periods of time. If hashrate is on a long-term upward trend, blocks will arrive faster than normal. Noticed how the 2012 block halving party was in November, but in 2016 it was in July? We will keep getting further ahead of schedule.
3/7 2. Even assuming constant hashrate, the timing isn't accurate. Blocks arrive 0.1% slower than intended. Half of that is due to an off-by-one bug in the implementation, half due to a probabilistic issue first discovered in my paper on predicting block halving party times.
4/7 3. It's very slow to react to changes in hashrate. This creates a positive feedback loop in some scenarios, such as a drop in price or competing chains with the same hash function, potentially leading to catastrophic failure.
5/7 It's not surprising that the algorithm is terrible - there's no indication that Satoshi knows anything about control theory, which is essential for designing systems like this.
6/7 Some alts have since experimented with other retarget algorithms. I've never really looked deeply into any of them, but perhaps there are some good ideas there.
7/7 Personally I'd support serious research into new retarget algorithms, with the goal of eventually implementing in Bitcoin. I might be holding a minority opinion, though.
